Memo: Learn Cloud Concept

Supanut Laddayam

8 min readApr 20, 2024

Cloud computing

is one practices to use a network of remote server hosted on internet to store, manage and process data.

characteristic of Cloud provider:

Someone else owns the servers.
Someone else hire IT People.
Someone else pay or rent the real estate.
You are only responsible for configuration cloud service and code.

Vs On-premise:

You own the servers
Your hire IT people
You pay or rent the real estate
You take all the risks.

The Evolution of Cloud Hosting

Dedicate server — One physical machine dedicate to a single business. That you purchase and put to off and set up networking, install OS and software

Benefits: you will get full control over server and have the highest level of security.

Use case: run a single web app

Pros: High security

Cons: Very expensive and highly maintenance

2. Virtual Private Server (VPS)— One physical machine dedicate to a single business. Isolate the application by virtualize machine to sub-machine to run multiple workload (application).

3. Shared Hosing — One physical machine share by hundreds of business. Isolate the virtual for each customer or tenant each other by folder the machine, so it will have 100+ business on single machine and you will limit the system work.

pros: Very cheap

cons: Very limited, can’t access to underlying OS, has some change the customer use over utilizing the machine and impact the workload

4. Cloud Hosting — Multi physical machine act as one system. To resolve the problem the one customer use over utilizing the single machine. The system is abstracted into multiple cloud services.

pros: Flexible, Scalable, Secure, Cost-effective, Highly configurability.

Cloud Service Provider

CSP — is a company that provide multiple Cloud service and those service can be chained together to create cloud architecture. eg. GCP, AWS.

*1st service of GCP is “App engine”.

7 Benefit of Cloud computing

Cost effective — you pay what you use (pay-as-you-go) with on-demand pricing, no up-front cost.
Global — launch workload to distributed anywhere in the world.
Secure — CSP take care of physical security. Service in cloud is secure by default.
Reliable — Data backup, disaster recovery (DR), data replicate and fault tolerance.
Scalable — Increase and decrease resource and service base on demand.
Elastic — Automate scaling during spike and drop in demand.
Current — underlying hardware and software is patched, upgraded and replaced by CSP without interrupt.

3 types of Cloud services

SaaS — software as a service, it’s a product is run and managed by service

provider. “Just use the application, don’t worry about the service is maintained”

usage: FOR CUSTOMER

2. PaaS — Platform as a service, it’s a platform that help you to deployment. and management the application. “Focus on deployment, Don’t worry about provisioning, config or underlying hardware or OS”

usage: FOR DEVELOPER

3. IaaS — Basic building block for cloud. provide access to network, compute and data storage space. “Don’t worry about IT staff, data center and hardware”

usage: FOR ADMINS

4 common cloud services for Infrastructure as a Service (IaaS)

Compute —having virtual computer that can run application, program and code.
Storage — having virtual hard drive that can store files.
Networking — having virtual network be able to define internet connection or network isolation
Databases — having virtual database for storing data.

Google’s Shared Responsibility model

shared responsibility model — is a visualization to determine the customer’s responsible and google’s responsible for related to GCP.

Customer’s responsible — about data and configuration of service and access control of user via permissions. “Responsibility in the cloud”
Google’s responsible — for the underlying infrastructure “Responsibility of the cloud”

Share Responsibility model for Google Compute service

IaaS

Bare Metal

Customer: Host OS config, Hypervisor

Google: Physical machine

Virtual machine (compute engine)

Customer: Guest OS config, Container runtime

Google: physical machine, hypervisor

Container (GKE)

customer: config container, deploy container and storage container

google: OS, hypervisor, container runtime

PaaS

Manage platform (App engine)

customer: upload code, config the env

google: server, OS, network, storage, security

SaaS

Google docs

customer: content of doc, mange file, config sharing access control

google: server, OS, network, security

FaaS

Cloud Function

customer: upload ur code

google: deployment, container runtime, network, storage, physical machine

SRM Overview

Customer:

IAM
OS
Network
Firewall
Client-side data encrypt
Server-side data encrypt
Networking Traffic protection
Customer Data

Google:

Compute
Storage
Network
Database
Region
Zone
Physical security

Cloud Computing Deployment model

4 models

Public cloud — aka Cloud Native: everything built on the CSP.
Private cloud — aka On-premise: everything built on the company’ data center.
Hybrid — using on-premise and CSP
Cross cloud — aka multi-cloud, hybrid-cloud: using multi CSP.

*Anthos is GCP service to handle to control across multi CSP and on-premise.

Cloud Computing Deployment models

Cloud — fully utilizing cloud computing

use cases: start-up, New project

2. Hybrid — using both cloud and on-premise

use cases: Bank, Legacy on-premise, Fin tech, Investment management

3. On-premise — deploy resource on-premise

use cases: Public sector (government), Sensitive data (eg. hospital), compliance and regulation (eg. insurance company)

Total Cost of Ownership (TCO)

the total expense and actual cost of IT spend.

2 type of cost:

CAPEX (on-premise) — spend money upfront for physical infrastructure:

software license, Phyiscal security, hardware, IT staff and maintenance.

server cost (compute), storage cost(hard drive), data center cost, techical staff

*with CAPEX: you have to guess upfront when you plan to spend.

OPEX (cloud / GCP) — pay for non-physical cost (on-demand / pay-as-you-go): leasing software, train employee, pay for cloud support, billing base on cloud metric (compute usage, storage usage)
with OPEX: you pay on service without investing in equipment.

Cloud Architecture Terminology

5 pillars

Availability — is ability to ensure a service remain available (HA),
Scalability — is ability to grow rapidly or unimpeded
Elasticity — is ability to shrink and grow to meed on-demand
Fault tolerance — is ability to prevent a failure
Disaster Recovery — is ability to recover from a failure

High Availability (HA)

ability for service to remain available by ensure there is no single point of failure and ensure a certain level of performance.

eg. system has load balancer and has compute distribute in 3 zone. if we lose one or two zone , we would still a running service.

Because load balancer will detect that there are unhealthy and reroute the traffic to remaining server -> it help us to be HA.

*Running workload across multiple zone -> ensure if 1 or 2 zone become available , your service / app remain avaialble.

High Scalability (HS)

ability to increase the capacity base on increase demand with rapidly of traffic, memory and compute power with 2 options:

Vertical scale — scaling up by increase the size of machine.
Horizontal scale — adding additional server with same size.

High Elasticity (HE)

ability to automatically increase or decrease the capacity base on current demand of traffic, memory and compute power.

GCP service: Manage Instance Group (MIG)

only for horizontal scale

scale out — add more server of same size.
scale in — remove more server of same size.

High Fault tolerance (HF)

ability for service to ensure there is no single point of failure -> prevent chance of failure.

“Fail over “— is when you have plan to shift traffic to redundant system in case primary system is fail.

GCP service: Cloud Load Balancing

eg. having a copy (secondary) of database where all ongoing change are synced. The secondary database / system is not in-use until a failover occurs and it become the primary database

High Durability (HD)

ability to recover from failure / disaster and prevent the loss of data.

solution that recover from disaster aka Disaster Recovery (DR):

Backup data ?
How fast can restore the backup ?
Backup is still work ?
How to ensure live data is not corrupt ?

Evolution of compute

Dedicated — a physical server (sole-tenant / single tenant, single node cluster, bare metal) utilized by a single customer.

pros:

Guarantee of security, privacy and full access underlying resources

cons:

Have to guess capacity
Overpay for underutilize server
Get upfront to purchase guess capacity
Can’t vertical scale -> need to manual migration
Limited by ur Host OS

VMs — a one physical machine that can run multiple Virtual machine.

add the software layer (Hypervisor) to handle to run the VMs
one physical server shared by multiple customer

pros:

Easy to Vertical or Horizontal scale
Easy to image for migration

cons:

Overpay for underutilized VM
Limited by ur Guest OS
Multi app on single VM can result in conflicts in resource sharing

Container — VM running multiple container, so has Docker Deamon layer that let multiple container run.

pros:

can maximize the utilize of available capacity
container share the same underlying OS -> containers are more efficient than multiple VMs
one container can run multiple without being limited and not conflicts during resource sharing.

Functions — are manage VMs running managed container aka Serverless Compute